Some Errors Found in Provided Dataset – Please Verify

[YiqunWang-FDU]

Hi @PTRRupprecht ,

Thank you for sharing your dataset and the accompanying Colab notebook for visualization and exploration.

While using your Colab script to explore the dataset, I noticed that there are some errors in the data. I have attached images highlighting the problematic data entries for your reference.

Could you please review and verify these issues in the dataset? Your confirmation and possible corrections would be greatly appreciated.

Thank you for your attention!

[PTRRupprecht]

Thanks for noticing! I will look into it!

[PTRRupprecht]

Hi @YiqunWang-FDU,

Thanks a lot for noticing these errors in the ground truth data!

First of all, I can confirm the errors that you spotted and was able to reproduce them with the online exploration tool and with additional offline scripts.

I looked into the issue and noticed a processing error (converting the ground truth data provided by the original publication to the Cascade format) specifically for a small subset of the DS21-jGECO1a dataset. Since the analyses in the original paper about Cascade were displaying results as median performances across neurons, these results were barely affect by the wrongly processed outliers.

I now fixed this issue and uploaded updated ground truth files for the two neurons where the problem occured (CAttached_Mohar16_jRGECO1a_V1_9_mini.mat and CAttached_Mohar16_jRGECO1a_V1_3_mini.mat).

Please note that the online exploration tool reads in the ground truth files in their order by time stamps, which means that the recording ID of the corrected recordings will be different compared to the recording IDs in your screenshots. In any case, the errors should now be corrected.

Thanks again for spotting this temporal alignment error, and let me know if you see any further problems!

Best,
Peter

[PTRRupprecht]

I'm closing the issue now, with the bug being fixed. If you feel that there is still something to be improved or to be discussed, please re-open the issue.

[YiqunWang-FDU]

Hi @PTRRupprecht ,

Thank you again for your prompt response and for correcting the previous errors in the dataset.

While continuing to work with the updated dataset, I have encountered another issue that appears to be an error in the data. I have attached images highlighting the problematic entry for your reference.

Specifically, I read your paper: Spike inference from calcium imaging data acquired with GCaMP8 indicators, where you mention that it can be challenging to identify single action potentials from noise when using older indicators like GCaMP6. However, in the trial shown in my attached image, there is a clear ΔF/F change visible, yet there is no corresponding electrophysiological recording of spikes.

Could you please verify whether this data entry is correct, or if there might be an issue with this particular trial? Your help in clarifying this would be greatly appreciated.

Thank you very much for your continued support!

[PTRRupprecht]

Hi @YiqunWang-FDU, thanks for spotting this apparent error! I will look into it and try to figure out where the problem lies!

[PTRRupprecht]

Hi @YiqunWang-FDU,

You were right, there was an issue with this specific trial of this recorded neurons, with the calcium imaging recording being much longer than the ephys recording. I fixed the issue by shortening the calcium imaging traces.

This issue affected only a small section of this specific neuron (and I did not find similar instances of such errors in other datasets so far), so the total effect on existing pretrained models is very small. Specifically the analysis from our recent preprint on GCaMP8 data and single-action potential detection is not affected since single action potentials in our analyses were detected based on electrophysiologically detected spikes (which are absent in this recording section that you detected).

If you find any other issue in the datasets, I'm happy to hear about it!

Best,
Peter

[YiqunWang-FDU]

Hi @PTRRupprecht ,

Thank you very much for your prompt response and clarification!

I appreciate your detailed explanation and the corrections you made. I’m also very interested in your research on spike inference, especially the identification of discrete spikes. Looking forward to learning more from your work!

Best regards,
Yiqun Wang

[PTRRupprecht]

Thank you! I'll close the issue for now.